python - How to replace LSTM with CuDNNLSTM

Question

I have built the following LSTM model however, it's very slow with huge dataset so, is it possible to replace it with CuDNNLSTM to gain GPU assistance? Here is my LSTM code:

def get_lstm_model():
    lstm_model = Sequential()
    lstm_model.add(LSTM(300, dropout=0.4, recurrent_dropout=0.4, input_shape=[1, 300], return_sequences=True))
    lstm_model.add(LSTM(64, recurrent_dropout=0.4))
    lstm_model.add(Dropout(0.5))
    lstm_model.add(Dense(1, activation='relu'))
    lstm_model.compile(loss='mean_squared_error', optimizer='rmsprop', metrics=['mae'])
    lstm_model.summary()
    return lstm_model

question from:https://stackoverflow.com/questions/65903731/how-to-replace-lstm-with-cudnnlstm

Answer 1

Waitting for answers